A year since Inuka's passing

It has been a year since Inuka, the first polar bear to be born in the tropics, passed away (or rather, was put to sleep). And with his passing, the Singapore zoo will no longer have a polar bear.

His passing brought up a lot of emotions in me. Not because I am a polar bear lover. But because of the significance of the polar bear enclosure to me.

You see, my father was involved in building the polar bear enclosure at the Singapore zoo. Specifically, the landscaping part. Going to the zoo with him (via the side entrance) when I was young was part of my school holiday routine. He went as part of work; I went under the reason of "helping him". Of course, I wasn't of much help, but it allowed me to roam the zoo for free.

More importantly, it allowed me to spend time with my father, seeing him at work, learning important life and leadership lessons from him.

Those times with him, those lessons from him, they are all the more precious because he passed away when I was 20.

And so, every time I pass by the polar bear enclosure (and the komodo dragon one too), I would remember the times with him, driving around Singapore as he visited various work sites to check on the progress of work, buy and bring lunch to his workers, and took over their work while they ate. He never gave any formal lectures on leadership to me, never talked about his work much, but I learnt from observing him. The importance of being present, seeing things with one's eyes, caring for the people in the team, and persevering through hard work. As a preteen and teenager, I was attending excellent leadership lessons for free.

Those lessons have served me well over the years, especially when I was in key leadership positions.

My son saw tears welling up in my eyes last year, when we walked through the empty polar bear enclosure at the zoo. I tried to explain to him the reason for those tears, but he might have been a bit too young to understand.

I only hope he will grow up learning as much from me as I had from my father.

Learning from AI

Paper: Visualizing memorization in RNNs

It is interesting to see how AI learns. What is especially interesting to me from this video and paper is the possibility of humans learning from machines. For example, in this example, the different AI implementations are trying to understand a piece of text, and we can see its "attention span" as it tries to understand a letter (and word). My thought: can this eventually be used to help us learn better in our own physical, human world? Can an AI technique be used to teach humans how to better read a sentence to grasp a better understanding of that sentence and its context? If so, it is an interesting example of how AI can help us to improve human skills, thereby contributing toward bettering human intelligence.

Update 4 on ESP32-based 3D printer control board

Good and bad news.

Good news: I got v1.0 of MRR ESPA (the basic 3D printer control board based on the ESP32) all soldered up. Motors with A4988 drivers working. Endstops working. Hotend was able to heat up to 240 degC and maintain it without any significant rise in the temperature of the MOSFET.

Bad news: The heat bed MOSFET smoked (almost started a fire) when I was testing it. Tried room temperature to 40 degC, and saw smoke when it started to heat up to 35 degC. Temperature on the heatsink attached to the MOSFET was reaching 120 degC. Shut off, then I tried gradual heating. Slowly went from room temperature to 35, then step up in steps of 2 to 3 degC. Eventually reached 50 degC. When going to 52 degC, there was a bit of smoke which soon stopped when temperature reached target. Then I tried 55 degC. That was when it started to smoke real bad. Like, things were melting, there was sizzling sound, and lots of smoke. Quickly shut off the mains.

You can see the black stuff that came from the MOSFET. Smoke was coming out, and when the smoke cleared, I saw that this black stuff had oozed out from the MOSFET.

This is the reverse side. The same blackish stuff was on the reverse side too, I removed it before taking the photo, though you can see some residue.

The heat bed is 180W, running on 12V, so the current draw (theoretically) is 15A. Resistance as measured from terminal + to terminal - for the heat bed was 0.9 ohms, which sounds about right. The MOSFET (IRLS3034) is driven by a voltage divider that drives the gate at 4.8V. Based on specs, the IRLS3034 should have Rds of 2.0 mOhms at 4.5V. So simple calculation says the MOSFET is dissipating 15x15x0.002 which is 0.45W. At temperature rise of 40degC/W, the MOSFET should only be rising like 18degC. I mean, even given other factors, the temperature rise should not be more than 50degC, right? But I was getting a temp rise of 100degC and more. I can't really figure out why.

If anyone has any idea, I would be open to hearing them.

Note: Issue fixed. Pre-launch version now available here. Some information on this has been updated here.

What "The Rise of Skywalker" may mean...

Star Wars: The Rise of Skywalker

Wow.

But what does it mean? There has been a lot of speculation.

So here, I add my own take.

The Star Wars movies made by George Lucas has been about Anakin Skywalker. Anakin Skywalker as Darth Vader in the original movies, then Anakin's story about how he came to be Darth Vader in the prequels.

And in the new movies, Episodes 7 and 8, Anakin's legacy through his children (Luke and Leia) in shaping the galaxy after the destruction of Death Star II and the collapse of the Empire.

The way I see it, the entire series, from Episode 1 till 9, is about Anakin Skywalker.

Who is Anakin Skywalker? Anakin Skywalker is the Chosen One. The one who will bring balance to the force.

And Kylo Ren has said that he will finish what Anakin has started. Which implies that Kylo will bring balance to the force. Or, he could be the one to put an end to Palpatine, a final end that Anakin started but did not accomplish (if Palpatine returns).

With the amount of things that George Lucas borrowed from Frank Herbert's Dune in the original movies, it would not be surprising that Episode 9 will continue to borrow some concepts from Dune. When Luke's voice said, "A thousand generations live within you," it reminded me about the Other Memory of the Bene Gesserit.

If that is so, there is nothing stopping Episode 9 from borrowing other concepts from Dune. Such as the warning against a central body (in the figure of Paul Atreides, and then the God Emperor Leto II) governing over all of humanity.

Could this mean that Anakin Skywalker, and the Skywalker line, is the force that seeks to prevent any one side, whether good or bad, from controlling the galaxy?

The rise of Skywalker, the balancing force. Could this be the meaning behind the movie title?

夜深难入眠

夜里独唱琵琶行
歌声未出泪先流
感慨万分不能停
泣梦九泉再重逢
转眼三年光阴逝
枯松独立泰山峰

--------------------------------------------
夜深星空难入眠
辗转反复心不静
意生字出句自成
只恨知音已无人

--------------------------------------------

穿山越岭孰不言

穿山越岭又曾见
赴汤蹈火入耳熟
赴汤蹈火又几人

--------------------------------------------
杜康不酿酒
羲之不提笔
太白不吟诗
孔明不谋计
志高技深无知音
青竹无名枉一生
玄德不顾孔明芦
卧龙再能无用处

 

--------------------------------------------

(Update 15 October 2022: A modification of the above)
杜康酿酒羲之书
太白吟诗孔明计
志高技深无知音
青竹无名枉一生
玄德不顾孔明芦
卧龙再能无用处

Determining what AI will replace

Real or artificial? Tech titans declare AI ethics concerns

I think we really need to sit down for a proper conversation on what we want to develop AI for. What is it that we want to replace with machines, and what is it that human beings will continue to do?

History has shown that if something is possible, human endeavor will eventually make it happen. Human beings can now fly, dive underwater, and go faster on land than any other land animal. We are now able to communicate across vast distances in real time. We can interact with people without having to be physically at the same location.

Still, human have been the ones at the center of these actions. Humans decide where to fly to, where to go diving, and where to travel on land. We decide on who we want to interact to, and what we want to talk about. But AI is slowly changing that. And if we do not come to a decision soon, AI will only make more and more decisions for us.

And the danger of that? If we make less and less decisions, our brains will grow stupid from lack of use.

Of course, this is not a danger if we can reuse that freed up brain power for something else. So the question is twofold.

1. What aspects of our lives are we willing to replace with machines, such that it frees us up from things that we do not (or cannot) do for us to focus on things that we should be really focusing on?
2. What are the things that we should be focusing on?

Looking at these two questions, I think we can say that machine vision is a good case for AI. It can help those who cannot see. It enhances our ability to see minute details so that we do not miss out on them. But at the same time, there is the downside: current machine vision is pre-trained before deployment, which means it will recognize a finite set of objects. Which means any reliance on such technology will "narrow" our perspective to that finite set, i.e. our attention gets drawn to things that the computer knows, making us more likely to miss other details that we might have otherwise caught if we did not rely on machine vision.

Let's look at machine transcription (aka speech-to-text synthesis). This definitely helps those who cannot hear. And for others, it frees up people from a mundane task so that we can use those people (resources) for other tasks. But is there a downside? So far, I can't think of any, but who knows?

Autonomous driving is another use case which has been undergoing heavy development. And a lot of attention in terms of ethics. Its controversy is: who is responsible should a self-driving car kill or injure someone? There is debate on this, and I hope people will take up this debate more actively, because with the current rate of development, self-driving cars are on the horizon. Let's sort the ethical questions before we are forced to face the reality of it.

And finally, my biggest fear: uncurbed, we will one day develop machines that are so intelligent, they no longer need human beings for their further development. They can develop themselves into even better and better versions able to do more and more things. Will we then find ourselves as second-class citizens of our own societies?

Do fifth-generation fighters require exceptional maneuverability?

The F-35's New OODA Loop

It is quite interesting that this article brought up COL John Boyd, the father of the OODA loop. It is interesting because COL Boyd was involved in designing the F-15 and F-16, and maneuverability was one of the main issues that he addressed. At a time when decision-makers were saying that missiles will end the era of dogfighting with guns, COL Boyd showed that fighters will still need be have high maneuverability in order to achieve their missions.

The question, then, is whether this still holds true?

"If superior performance is not required to win the next war, what is? What characteristics will allow a fifth-generation fighter pilot to execute the OODA loop at a faster tempo than his enemy in a combat scenario he is likely to encounter? The answer, in a word, is information."

Yes, the OODA loop places heavy emphasis on information, because information is crucial for the orientation process of the loop. But the entire process is not sequential; everything is interacting with each other, and a fighter's maneuverability affects the orientation, decision-making, and action processes too.

"While maneuverability allowed a fourth-generation fighter pilot to operate the OODA loop at a faster tempo than his enemy in a dogfight, this is not the primary conflict a fifth-generation fighter is designed around. Today’s designers optimized the F-35 to execute a coordinated OODA loop between a flight of F-35s performing an air-to-air or air-to-ground engagement with today’s weapons and tactics, not a 1970s one-on-one dogfight."

So while today's fighters may not need to be as maneuverable because maneuverability is being designed into missiles instead, we need to ask the question: can the enemy outmaneuver our missiles? Because if the enemy can do so, it can close the distance to our fighters, and we will still end up in a dogfight. And we will end up in a dogfight against an enemy that is designed for dogfights using a fighter that is not. Who will win?

That same question came up in COL Boyd's mind back in the 60s and 70s, when MiGs came up against F-4s, designed to fight with missiles, in the Vietnam War.

COL Boyd's emphasis on maneuverability is about winning under all circumstances. The execution of the OODA loop must be faster than the enemy's under all circumstances. That was why he came up with the E-M charts to show the performance of various aircraft, to illustrate where their advantages and disadvantages are. That was why he sought to emphasize the design of maneuverability into fighter design.

While information is key to winning battles, maneuverability is an equally important aspect that must not be neglected. As long as human beings are the ones fighting wars, we need to be able to gather information faster than our enemies, orient ourselves faster, decide faster, and be able to act faster.

Under all circumstances.

Should we hedge our bets and emphasize superior information, superior decision-making, and superior maneuverability all at once? Or do we place a gamble that superior information is all we need?